10 1 mioko saito

－ 165－

CICE Hiroshima University, Journal of International Cooperation in Education, Vol.10 No.1 (2007) pp.165～ 182

Construction and Application of SACMEQ School Resources:Portray of School Systems

Based on the Rasch Scaling Technique

Mioko SaitoUNESCO International Institute for Educational Planning, France

AbstractA School Resources Index (SRI) was constructed using the Rasch scaling technique.

Data were from the pooled school resource items that were collected from pupils,

teachers, and school heads derived from two sub-regional studies undertaken by

Southern and Eastern Consortium for Monitoring Educational Quality (SACMEQ)

in 1995 and 2000. Capitalizing on the possibility to link more items, the new scale

provided a stronger positive correlation between the school resources and the pupils’

achievement compared to using the initial variable containing the summation of

school resource possessions. This SACMEQ SRI provided a reliable and valid

instrument to compare the resource level among different school systems with varied

levels of economic development. It also provided a meaningful tool to examine the

changes in the school resources between two different time points. From the SACMEQ

SRI, the hierarchical resource “profiles” were also established. The profiles could be

used as a guideline of standard to identify more critical and relevant resource items

at each progress level, for better budgetary planning, resource allocation, priority

setting, and benchmarking.

Introduction

Policy concerns and research questionsThe economic difficulties and financial constraint during the 1980s and 1990s pressured

many policy makers to look into educational productivity encompassing more efficient useof educational resources. Specifically they are concerned about: (i) whether the schoolinvestment improves the educational quality; (ii) which resources are effective to education;and (iii) the minimum requirement on school resources.

These policy concerns were also the key issues of the educational policy researchproject undertaken by a consortium known as Southern and Eastern Africa Consortium forMonitoring Educational Quality (SACMEQ). SACMEQ has produced a data archive in2004 (Ross et al. 2004) which contained data from two large scale sub-regional studiescovering over 60,000 Grade 6 pupils and some 7,000 teachers in over 3,000 schools inBotswana, Kenya, Lesotho, Malawi, Mauritius, Mozambique, Namibia, Seychelles, SouthAfrica, Swaziland, Tanzania Mainland, Uganda, Zambia, Zanzibar, and Zimbabwe.

In order to address these policy planning issues, it has been decided to measure the

Mioko Saito

－ 166－

pooled school resource items in a meaningful way. In this exploratory study, the majorattempt has been made to establish an index of school resources that is valid across fifteenSACMEQ countries over time. In particular, the following six research questions are beingaddressed:

● Which school resource items constitute the SACMEQ SRI?● What are the distinctive profiles of the schools at different resource levels on the

scale?● What is the breakdown of schools across the resource levels for each SACMEQ

country?● What is the mean school resources score for each SACMEQ country?● What are the changes in the school resources between SACMEQ I and SACMEQ

II; and● What are the among-region and within-region variations of the school resources

scores in each SACMEQ country?To test the utility of the final scale, a further research question has been added:

● What is the relationship between school resources and pupils’ achievement outcomeson reading and mathematics in SACMEQ countries?

Rational for constructing a School Resources IndexWhen measuring the resource level of a school, one particular resource item may not

provide a reliable indicator, just as one single mathematics question is not sufficient todefine a mathematical competence. In a typical School Survey Questionnaire, questions onschool resources maybe inter-related. For example, a school library is associated with somebooks to borrow and bookshelves to hold books. Furthermore, some items are more difficultto fulfil than others. For instance, electricity is a condition for having electrical appliances.

A scale should be appropriate for the developmental level of a country. If a list ofschool resource items in a School Survey Questionnaire for an industrialized country wasused in a developing country, it may be criticized for the contextual irrelevance and themagnitude. For example, a relevant question used in a developing country asking whetherthey have water at school may not well discriminate schools in an industrialized country.Likewise, while a question about computers in a developing country may not be relevant, aquestion which really discriminates schools in an industrialized country would be askingthe magnitude of the item. It is therefore critical to combine these questions in order tomeasure countries with various development levels on a single continuum.

Scope and limitationsIn this study, the Rasch measurement approach has been used to construct the SRI.

Various research studies explore the uses of different Item Response Theory (IRT) modelswith two or three parameters on item characteristics (Write & Stone 1979). However, it isnot the intention of the study to compare the Rasch model against other IRT models.

The focus of this study is the availability of materials and physical resources. Other

Construction and Application of SACMEQ School Resources: Portray of School Systems Based on the Rasch Scaling Technique

－ 167－

Pe

evi

ßv

v i=

+

−

−

δ

β δ

1

1

school variables were not included in this study. However, two areas of information requirefurther clarification. First, the pupil- and the school-level questions about whether pupilscan borrow library books were considered as availability of “borrowable” books. Secondly,questions on the distance from the school to facilities were regarded as “alternative access togoods” in the vicinity, because of the importance of this potentiality in developing countries(Griffin 2004a).

In the present study, the demonstrated relationship between school resources and pupilachievement does not adjust for the pupil input level. It is argued that the relationship betweenpupil achievement and the school resources using the newly established scale is strongerthan using the existing derived composite variable as a result of the possibility of incorporatingmany more items. The aspect of measuring the impact of school resources on pupilachievement when controlling for the pupil home background shall be dealt with in otherstudies.

Literature Review

Rasch measurement modelWith the classical testing method, as opposed to the IRT, it would be difficult for a

given score to sufficiently have any substantial meaning as regards to the fulfilment abilityof the respondents. This is attributed to the fact that the scores are dependent on the sampleof respondents and the sample of items in an instrument. For example, an average respondentis expected to score high on an “easy” instrument and low on a “difficult” instrument (Woo2005a).

The Rasch model is a probabilistic measurement model (Rasch 1980; Write & Stone1979) that is the simplest and the most commonly used within the existing models of IRT.The Rasch model incorporates one parameter, namely item location, for modelling the itemand respondents’ behaviours. Using this model, the school resource item difficulty and theschool’s fulfilment ability for a school resource item can be placed on one single scale.

Given the school’s resource location (ßv) and the resource difficulty (δi), the probabilityof a school (v) to fulfil a resource item (i) can be expressed as:

A given score in the Rasch measurement can provide a probabilistic description of theresources that a school has. Given a total score of r, it is most likely that the r resource itemsa school has are the ones which are easiest on the scale.

In addition, in the Rasch approach, an assumption has been made in which the singlescale that maps the schools and the resource items is sample-free, but not population-free(Griffin 2004b). That is, within a set of resources and a set of schools for which the modelholds, subsets of resource items give the same relative school locations, but with differentprecision depending on the number of items. In this sense, the measure of school resources

Mioko Saito

－ 168－

is independent of the instrument used to measure the characteristics of the school. Thus oncondition that the responses fit the model, it is possible to link inter-related subsets of itemsfrom different data collections to make a meaningful scale. Finally, some ordered responsecategory items of the Likert form can be also identified with a meaningful hierarchy withinthe same scale (Andrich, de Jong & Sheridan 1997).

Based on the IEA Reading Literacy data on 9-year and 14-year old students, Siniscalco& Ross (1997) carried out an experiment to establish an international reading resourcesscale using IRT. They reported that it was possible to construct a meaningful resource scalein order to map 30 countries.

School resourcesResearch on the relationship between school inputs and pupils achievement has been

controversial. At foremost, the Coleman study (Coleman et al. 1966) shocked the educationalpolicy makers as it reported that the pupil’s socio-economic background was more influentialthan the school inputs on the students’ performance. Subsequently, Peaker (1971) concludedthat student background and early learning do matter on the progress of achievement, notthe school effect.

During the 1980s and 1990s, these findings were greatly challenged in a various studiesin the context of non-industrialized countries. (Heyneman 1980; Heyneman & Loxley 1983;Fuller 1985; Fuller 1987; Hanushek 1995; Hanushek 1997). The conditions that wereestablished prior to schooling in developing countries were not as influential determinantsof achievement as in high-income countries. The researchers criticized that the variablesthat were often used in industrialized countries to measure student background were notappropriate in developing countries. Consequently, there was not much variation in thesevariables.

There exist a number of other studies conducted (Murimba et al. 1997; Varghese 1995;Fuller & Clarke 1994; Harbison & Hanushek 1992; World Bank 2004; Postlethwaite &Ross 1992). In general, these studies reported that facilities, textbooks, and writing materialshad reasonable effect on the student performance in developing countries. In the ThirdInternational Mathematics and Science and Study (TIMSS), Martin et al. (1999) reportedthat the shortage of instructional materials would affect at a different magnitude in developingcountries as opposed to in industrialized countries.

Methodology

Population and sampleThe target population for SACMEQ studies was defined as “all pupils at the Grade 6

level during the data collection year at the eighth month of the school year who were attendingregistered government or non-government schools in the country”. In some countries, smallpercentages of population were excluded due to schools affected by wars and natural hazardsand schools using non-mainstream curriculum.


－ 169－

The sample was selected in two stages. At the first stage, a sample of schools wasdrawn with a probability proportional to the enrolment on Grade 6. The intra-class correlation(Rho) values obtained in the SACMEQ I study were used where possible. Otherwise, a Rhoof 0.4 was used for SACMEQ II. The accuracy required for all school systems was set at theequivalent of a simple random sample of 400 pupils, which would yield a sampling error of5 percent for a percentage and 0.10 of a standard deviation, with a confidence limit of 95percent. At the second stage of sampling, a simple random sample of 20 pupils from allGrade 6 pupils was drawn within each selected school.

Sampling weights were applied in order to adjust for (i) discrepancies between thepopulation at the time of sampling and the data collection; (ii) disproportion among strata,and (iii) differences between the planned and achieved samples within strata. The designeffect obtained via the IIEP JACK (Ross & Leite 2000) demonstrated that the sample designsin SACMEQ I and II had satisfied the prior requirements of sampling accuracy, except forSouth Africa and Uganda, due to the under-estimation of the Rho value at the time of sampling.

A more detailed account of the sampling procedures used in the SACMEQ studies hasbeen presented in the SACMEQ data archive (Ross et al. 2004).

Data preparation(a) Item linkage

As shown in Table 1, the items concerning school resources included 64 SACMEQ Iitems (10 items from the Pupil Questionnaire, 15 items from the Teacher Questionnaire, and39 items from the School Head Questionnaire) and 64 SACMEQ II items (13 items fromPupil Questionnaire, 15 items from the Teacher Questionnaire, and 36 items from the SchoolHead Questionnaire). Out of the 77 total items, 51 items were common to both studies,which were used as “anchor items” linking the two studies.

(b) Data matrixThe Rasch Unidimensional Measurment Models (RUMM) software (RUMM

Laboratory 2003) was used in order to calibrate the school resource items. The valid caseswere aggregated at the school level. Overall, 2890 schools were used for the calibrationprocess. The country, the school location, and the time were used as “factors”.

(c) Recoding of variablesQuestions regarding the availability of items and questions on the quantity of items

were recoded to have dichotomous values, i.e., “not available (0)” or “available (1)”. Questionson the distance to facilities have been recoded as “over 5 kilometres (0)” versus “up to 5kilometres (1)”. The items with ordered response categories were recoded to have threecategories. Specifically, the textbook questions were recoded as “no textbook (0)”, “sharetextbook (1)”, and “have own textbook (2)”. The question on building condition was recodedas “require rebuilding (0)”, “require some repairs (1)”, and “in good condition (2)”.

(d) Item calibrationWith RUMM’s “conditional maximum likelihood” procedure, the school parameters

were eliminated while estimating the resource parameters. Based on the resource parameter

Mioko Saito

－ 170－

IIQEMCASIQEMCASsnoitseuQ- Exercise books PQ22.1 PQ21.1

- Notebooks PQ22.2 PQ21.2

- Pencils PQ22.3 PQ21.3

- Rulers PQ22.4 PQ21.6

- Pencil erasers PQ22.5 PQ21.5

- Ball point pens PQ22.6 PQ21.7

- Pencil sharpeners PQ21.4

How many of the following items do you have

this term?

- File folders PQ21.8

- Reading textbooks PQ20 PQ35How are the textbooks used in your

classroom?- Mathematics textbooks PQ38

Are you allowed to take library books home

from school?PQ21 PQ20

What do you sit on in your classroom? PQ24 PQ22

What writing place do you have in your

classroom?PQ25 PQ23

How many books do you have in your

classroom library or book corner?TQ8 TQ10

- A usable writing board TQ10.01 TQ12.1

- Chalk TQ10.02 TQ12.2

- Wall chart TQ10.03 TQ12.3

- A map of your country TQ10.04

- A map of Africa TQ10.05

- A world map TQ10.06

- Cupboard TQ10.07 TQ12.4

- Bookshelves TQ10.08 TQ12.5

- Classroom library, book corner or book box TQ10.09 TQ12.6

- A water tap TQ10.10

- A teacher table TQ10.11 TQ12.7

- A teacher chair TQ10.12 TQ12.8

- An atlas TQ10.13

Which of the following are available in your

classroom or teaching area?

- An English dictionary TQ10.14

- A map TQ13.1

- An English dictionary TQ13.2

- Geometrical instruments TQ13.3

- Teacher’s guide (English) TQ13.4

Which of the following do you have access to

in your school?

- Teacher’s guide (Mathematics) TQ13.5

What exactly have you used the education

resource centre for during the academic year?TQ24

- Health centre / clinic SQ11.1 SQ13.1

- Tarmac road SQ11.2 SQ13.2

- Public library SQ11.3 SQ13.3

- Book shop SQ11.4 SQ13.4

- Secondary school SQ11.5 SQ13.5

How many kilometers is it by road from your

school to:

- Shopping centre or market SQ13.6

- Permanent classrooms SQ27.1 SQ34.1

- Temporary classrooms SQ27.2 SQ34.2

How many teaching areas does your school

have?

- Open-air teaching areas SQ27.3 SQ34.3

- Permanent SQ28.1 SQ35.1 What is the total inside area of all permanent

and temporary classrooms in your school?- Temporary SQ28.2 SQ35.2

Table 1. List of Variables Considered for the First Exploration of SACMEQ SRI


－ 171－

Mioko Saito

－ 172－

Table 2. Summary of Calibrations

estimates, the school parameters were estimated using a “direct maximum likelihood”procedure (Andrich & Luo 2003). One of the fit statistics generated was a standardizedresidual with a theoretical mean of 0 and standard deviation which approaches 1. It wasprovided for both schools and resources.

Several calibration runs were executed in order to come to the final two versions of theschool resource scale: one with 72 items and another with 68 items. Table 2 shows thesummary. A more detailed account of the item calibration procedures has been presented inSaito (2005).

(e) Scoring all schools using the item parametersAll the schools that have participated in SACMEQ I and/or SACMEQ II were scored

using the calibrated item parameters. This was undertaken by RUMM’s “person measure”procedure. The school estimates were scored using both the 72 item-scale and the 68-itemscale. The scores were then merged back to the original combined pupil-level data file forfurther analyses. The correlation coefficient between the 72-item scale and 68-item scalewas 0.991. The results shown below were based on the 72-item scale.


－ 173－

Results

School resource scale from the sample schools in the calibrationThe reliability on the person separation index with the 72-item scale was 0.901,

providing the power of test-of-fit rating as “Excellent”. While the mean of the item locationwas fixed as 0 by definition, the mean of the school location was 0.133, and the standarddeviation was 1.058. The mean of 0.133 indicates that the average school in SACMEQcountries was more “able” than the “difficulty” level of the average resource item.

A response made by a school to each item can be considered as a function of twoforces: (1) the intensity of an item; and (2) the fulfillment of a school. The probability that aschool has a given school resource is a function of the level of “resource power” of the itemand of the level of “resource fulfillment” of the school (Siniscalco and Ross 1997). It mustbe noted, however, that this generalization is a “probabilistic” one.

In Figure 1, the individual resource items’ thresholds and the distribution of schoolsaccording to their resource level have been placed on the 72-item scale. A resource itemappearing at a higher level would exist in fewer schools than the item appearing at a lowerlevel. For example, a duplicator (location 1.131) is likely to exist in fewer schools than aschool library (location 0.071) would. On the other hand, a teacher table (location -0.945)would exist in more schools than a school library would.

Also shown in Figure 1 is the school position in relation to the individual resourceitems. For example, an average school in Seychelles (location 2.104) is likely to haveeverything except some “high-tech” equipment (location above 2.216). There is a 50-50chance of having school buildings in good condition (location 2.19). On the other hand, anaverage school in Malawi (location -0.871) is likely to fulfil all the lower items up to ateacher table (below -0.945) and have a 50-50 chance of fulfilling “the distance from theschool to the nearest market within 5 kilometres” (location -0.849).

Portrait of six school resource levelsUsing the information in Figure 1, it was possible to develop a portrait of six typical

schools with increasing levels of school resources. These were based on the identification ofthe common characteristics in groups as well as the evaluation of the location estimates andthe overall distribution. Some systematic hierarchy from independent questions appeared inthe levels. For example, a form of water appeared in Level 2, piped water in Level 3, andwater in classroom in Level 6. Electricity showed up in Level 4, simple appliances inLevels 4 and 5, and high-tech appliances in Level 6.

(i) Level 1: Insufficient School ResourcesA school at this level has a playground. Buildings require repairs. Classes take place in

an open-air class. There is a writing board and chalk. Pupils have or share sitting/writingplaces. However they share textbooks. Pupils have at least an exercise book, a ballpoint pen,and a pencil.

Mioko Saito

－ 174－

Figure 1. Item Map for SACMEQ School Resources


－ 175－

(ii) Level 2: Limited School ResourcesIn addition to the above resources, a school has a clinic and a market within 5 km from

the school. It has a sports ground and non-piped water. Classroom has a temporary and/orpermanent structure. The school head has his/her office. In the classroom, there is a chairand a table for the teacher. Teachers have an access to an English dictionary in school.Pupils have rulers.

(iii) Level 3: Basic School ResourcesIn addition to the above resources, a school has a fence, a store room, and a staff room.

Water is piped, and a toilet is used by less than 60 pupils. In the classroom, there is a wallchart. Teachers have access to a map and teaching guides. They have at hand Englishdictionary, map of a country, and an atlas. Pupils have at least a notebook and an eraser.

(iv) Level 4: Comfortable School ResourcesIn addition to the above resources, a school has electricity and a school library, a

classroom library, and a first aid kit. It has a radio, telephone, and a typewriter. In theclassroom, there is a cupboard. Teachers have world map, map of Africa at hand in theclassroom, and they also have an access to geometric instruments. Pupils have their owntextbooks, and sharpeners. The school purchases library books every year, and books can beborrowed.

(v) Level 5: Affluent School ResourcesIn addition to the above resources, a school is located within 5 km from a public

library and a bookshop. A water tap can be found in a classroom. A school has a duplicatorand a tape recorder. In the classroom, there are bookshelves. At least one book per pupil isavailable for both class and school libraries. Pupils have file folders.

(vi) Level 6: Prosperous School ResourcesIn addition to the above resources, a school has a secretary’s office; a school hall and

a cafeteria. School buildings are in good condition. The teaching space is at least 2m2 perpupil. The school has a TV, a VCR, a photocopier, a computer, an overhead projector, a faxmachine, and a film projector.

Distribution across the hierarchical SACMEQ school resource levelsThe percentages of Grade 6 pupils who were in schools with different resource levels

along with the standard errors of sampling (SE) have been presented in Table 3. It is shownthat the distribution was negatively skewed in Mauritius and Seychelles (SACMEQ II). Onthe other hand, large percentages were found on lower levels for Malawi and Zanzibar(SACMEQ I). On the whole, the category which yielded the highest percentage for SACMEQI was Level 2 (31 percent) where as for SACMEQ II, it was Levels 3 and 4 (28 percent).

Mioko Saito

－ 176－

Table 3. Percentage and Sample Error for Each School Resources Level

ZSRESLOCEstimate

= ×+

+1000 09215172409658

0 9420927535299500

.

.

Mean school resourcesEach school has been given an estimate of a school resource score. When merged back

to the pupil-level original data file, the mean school resource score (at pupil level) was-0.09215172409658 and its standard deviation was 0.960755343058. The mean schoolresource should be interpreted as the school resource level that pupils have access to. Astandardized school resource score (ZSRESLOC) has been calculated making the SACMEQII mean as 500 and the standard deviation as 100 as shown in the following formula:

Where Estimate is the individual school resource score calculated at pupil level.

In Table 4, the means and standard errors of sampling for the school resources for eachcountry using the standardized score have been presented. For SACMEQ I, the overall meanschool resource score was 464.8. The mean ranged from 379.3 in Malawi to 630.8 in


－ 177－

Table 4. Mean and Sample Error for School Resource Scale

Mauritius. For SACMEQ II, the country with the highest mean was Seychelles (675.2), andthe lowest mean school resource score was 409.7 in Malawi.These results were consistentwith the level of GDP per capita (World Bank 2003).

Changes in the school resource level between SACMEQ I and SACMEQ IIAlso shown in Table 4 are the differences between two studies. The standard error of

the difference was calculated by taking the square root of the sum of the variances for eachmean. In order for the difference to be significant at the 95 percent confidence level, thedifference must be greater than or equal to two standard errors.

All the countries except Mauritius increased the school resources. The increased valuesvaried from 12.6 points in Zambia to 32.8 points in Namibia. The changes in Kenya, Malawi,Namibia, and Zanzibar were statistically significant at the 95 percent confidence level takinginto consideration the size of the standard errors.

Mioko Saito

－ 178－

Table 5. Variation Among and Within Regions

ARVF

Fm

n

=−

+ −

1

1

Among-region and between-region variationsThe distribution equity was examined using the allocation patterns of the school

resources. In Table 5, the variation among and within regions on the school resources havebeen presented for each school system. The variations among regions (ARV) were calculatedusing the F statistics obtained from the ANOVA analyses conducted at the school-level datafile. The following formula was used:

Where m = number of schools; n = number of regions

For SACMEQ I, Malawi, Mauritius, and Zambia showed small variations amongregions (less than 0.1). However, in Kenya and Namibia, these figures exceeded 0.45. ForSACMEQ II in Mauritius and Uganda, the calculated variations among regions were negative,which were treated as no variation among regions. In contrast, in Namibia and Zanzibar, thevariation among regions yielded large figures. When comparing SACMEQ I and II, in Kenyathe variation among regions reduced by 0.14 points where as in Namibia and Zanzibar, theyincreased by about 0.11 or 0.13 points.


－ 179－

Table 6. Coefficient of Correlation between SACMEQ SRI and Reading andMathematics Scores of Pupils in the SACMEQ I and SACMEQ I

Correlation between the SACMEQ school resource scale score and the achievement scoresThe SACMEQ Reading and Mathematics scores in the SACMEQ archive had been

established using the Rasch scaling method (Ross et al. 1994; Andrich et al. in press), andstandardized to have a mean of 500 and standard deviation of 100 for SACMEQ II. It shouldbe noted that the overall correlation between the existing simple resource index (summationof 22 items) and the aggregated achievement scores at school level for SACMEQ II were.446 and .388 for Reading and Mathematics respectively. However, when using the newly-established resource scale, the correlation improved to .490 and .437 respectively. This wasdue to the fact that the Rasch model enabled more items to be incorporated to build thescale.

The coefficients of correlation between the SACMEQ school resources score and Readingand Mathematics scores for each country have been presented in Table 6. In general, theachievement scores in both subjects were positively related to the school resources score withstatistical significance. This finding was consistent with the conclusion of Siniscalco and Ross(1997). The magnitude of the correlation coefficient was particularly high in Kenya, Namibia,and Zimbabwe (SACMEQ I), and Botswana, Namibia, and South Africa (SACMEQ II).

In order to put perspective to the magnitude of these correlation coefficient, it shouldbe noted that the Reading Literacy Study conducted by the IEA reported the school-levelcorrelation coefficient between school resources and achievement ranging from -0.03 to+0.14 (Ross & Postlethwaite 1994).

However, these relationships were not adjusted for the pupil input. Further studiesneed to be carried out in order to determine the veritable relationship.

Mioko Saito

－ 180－

Conclusion and Discussion

The established SACMEQ SRI satisfied criteria of an “ideal” scale (Woo 2005b). Firstof all, the SACMEQ SRI exhibited an “excellent” reliability index of 0.901, indicating thatthe scale is internally consistent and is separating the schools on this variable. The highreliability could be due to the dependency of items within the scale. For example, manyhigh-tech items were dependent on the availability of the electricity. Using the Rasch model,it would be worthwhile to re-examine the reliability and the fit of the scale by converting thedependent items to a single item with a greater total score with minimized redundancy.

Secondly, a given score of a school on the school resource scale was not merely aquantitative measure. It provided a qualitative description of the resource items that arelikely to be available in a school. These profiles could be used as a guideline to identifycritical and relevant resource items at each progress level.

Thirdly, the SACMEQ SRI had a correlation coefficient of 0.490 and 0.437 with theschool aggregates of pupil Reading and Mathematics scores respectively, suggesting thatthe scale can be used as a predictor of school achievement as well as an indicator of aschool’s developmental level. This was achieved by a careful selection of the resource itemsthat were meaningful in developing countries. A similar approach would be required inorder to validly measure the socio-economic status of pupils in this context (Dolata 2005).This would shed light on the long-lasting debate regarding the interaction between schooleffect and the pupil background on achievement.

Finally, using the linked common items between SACMEQ I and SACMEQ II, countriesthat only participated in one of the studies were also placed on the same scale as those thatparticipated in both studies. However, interchanging totally one set of items for another setwould significantly alter the nature of the resource list. That is, very different subsets ofschool resources would mean different qualitatively different functioning in schools. Thesituation can be accepted if each item on the resource instrument is taken to represent theuniverse of that type of resources. For example, a blackboard is an item representative of allpossible types of blackboard, and a copier is a representative of copiers, etc. Such anassumption does not allow that the copier item is exchanged for a blackboard. Such a changein the instrument would shift the meaning of the overall resource allocation and would makethe measure of resources dependent on which items were used on the instrument and wouldviolate the assumption of specific objectivity. However, if only a few of the resource itemsare different among the schools, then depending on which ones they are, they might beexchangeable in the above sense. This notion of exchangeability may require furtherexploration in order to justify the use of Rasch in the domain of school resources.


－ 181－

References

Andrich, D. & Luo, G. (2003). Conditional estimation in the Rasch model for ordered response

categories using principal components. Journal of Applied Measurement, 4, 205-221.

Andrich, D., de Jong, J. H. A. L. & Sheridan, B. E. (1997). Diagnostic opportunities with the

Rasch model for ordered response categories. In J. Rost and R. Langeheine (Eds.), Applications

of latent trait and latent class models in the social sciences (pp. 59-70), Munster/New York:

Waxmann.

Andrich, D., Luo G., Ross, K., Saito, M. & Dolata, S. (in press) Analysis of the Southern and

Eastern Africa Consortium for Monitoring Educational Quality (SACMEQ) data. Paris:

UNESCO/IIEP.

Coleman, J. S., Cambell, E., Hobson, C., McPartland, J., Mood, A., Weinfall, F. & York, R.

(1966). Equality of educational opportunity (Coleman) study (EEOS). Washington, DC:

Department of Health, Education and Welfare.

Dolata, S. (2005). Construction and validation of a pupil socioeconomic status index for SACMEQ

education systems . Paper presented at SACMEQ Conference.

Fuller, B. (1985). Raising school quality in developing countries: What investments boost learning?

(Report No. EDT7). Washington, DC: The World Bank.

Fuller, B. (1987). What school factors raise achievement in the third world? Review of educational

research, 57 (3), 255-292.

Fuller, B. & Clarke, P. (1994). Raising school effects while ignoring culture? Local conditions

and the influence of classroom tools, rules, and pedagogy. Review of educational research, 64

(1), 119-157.

Griffin, P. (2004a). Chapter 5 Student characteristics. Unpublished document.

Griffin, P. (2004b). Chapter 3 The development of the teacher tests, equating pupil and teacher

results, and international benchmarking. Vietnam reading and mathematics assessment study

Volume 3. Hanoi: The World Bank.

Grouws, D. A. & Cebulla, K. J. (2000). Improving student achievement in mathematics.

Educational Practices Series 4. Brussels: International Academy of Education and UNESCO

Geneve: International Bureau of Education.

Hanushek, E. A. (1995). Interpreting recent research on schooling in developing countries. The

world bank research observer, 10 (2), 227-246.

Hanushek, E. A. (1997). Assessing the effects of school resources on student performance: An

update. Educational evaluation and policy analysis, 19 (2), 141-164.

Harbison, R. W. & Hanushek, E. A. (1992). Educational performance of the poor: Lessons from

rural Northeast Brazil. Washington, DC: The World Bank.

Heyneman, S. P. (1980). Differences between developed and developing countries: Comment on

Simmons and Alexander’s “Determinants of school achievement”. Economic development

and cultural change, 28 (2), 403-406.

Heyneman, S. P. & Loxley, W. A. (1983). The effect of primary-school quality on academic-

achievement across 29 high-income and low-income countries. American Journal of Sociology,

Mioko Saito

－ 182－

88 (6), 1162-1194.

Martin, M. O., Mullis, I. V. S., Gonzales, E. J., Smith, T. A. & Kelly, D. L. (1999). School

contexts for learning and instruction. Chestnut Hill: TIMSS International Study Centre, Boston

College.

Murimba, S., Moyo, G., Pfukani, T. M. & Mtembo, R. (Eds.) (1995). The analysis of educational

research data and policy development: an example from Zimbabwe. International Journal of

Educational Research, 23 (4), 305-383.

Organization for Economic Co-operation and Development (2004). Learning for tomorrow’s

world – First results from PISA 2003. Paris: OECD.

Organization for Economic Co-operation and Development (2005). School factors related to

quality and equity. Paris: OECD.

Peaker, G. F. (1971). The Plowden children four years later. London: National Foundation for

Educational Research in England and Wales.

Postlethewaite, T. N. & Ross, K. N. (1992). Effective schools in reading: Implications for

educational planners. The Hague: The International Association for the Evaluation of

Educational Achievement.

Rasch, G. (1980). Introduction. Probabilistic models for some intelligence and attainment tests.

(pp. 3-12), Chicago: The University of Chicago Press (original work published in 1960 by the

Danish Institute for Educational Research).

Ross, K. & Postlethwaite, T. N. (1994). Differences among countries in school resources and

achievement. In W. B. Elley (Ed.), The IEA study of reading literacy: Achievement and

instruction in thirty-two school systems (pp. 123-147), Oxford: Pergamon.

Ross, K., Saito, M., Dolata, S., Ikeda, M. & Zuze, L. (2004). Data archive for the SACMEQ I

and SACMEQ II projects. Paris: IIEP-UNESCO.

RUMM Laboratory (2003). RUMM 2020 getting started. Murdock: RUMM Laboratory Pty

Ltd.

Saito, M. (2005). The construction of a “SACMEQ school resources index” using rasch scaling.

Paper presented at the SACMEQ Conference (28 September 2005), Paris, France.

Siniscalco, M. T. & Ross, K. N. (1997). The establishment of an international reading resources

scale: An exploratory study using modern item response theory. Paris: UNESCO/IIEP.

Varghese, N. V. (1995). School facilities and learner achievement: Towards a methodology of

analysing school facilities in India. Perspectives in Education, 11 (2), 97-108.

World Bank. (2004). Vietnam: Reading and mathematics assessment study, Volume 2. Hanoi:

World Bank.

World Bank. (2003). World development indicators database. Washington, DC: The World Bank.

Woo, M. (2005a). An ideal measurement scale. Unpublished document.

Woo, M. (2005b). Classical test theory. Unpublished document.

Write, B. D. & Stone, M. H. (1979). The measurement model. Best test design: Rasch measurement

(pp. 1-17). Chicago: Mesa Press.

10 1 mioko saito

Education

index of school resources

school resources indexwhen

mean school resources

school systemsbased

school heads

school resourcesscores

school library

school resources index